Multilingual MMLU - Bengali: leaderboard
The "Bengali" subtask of Multilingual MMLU. A language split of Multilingual MMLU, measuring multiple-choice accuracy on translated MMLU questions across academic and professional subjects.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 74.63 | #350 |
| 2 | Qwen 2 72B Instruct | 68.26 | #576 |
| 3 | Qwen 2.5 72B Instruct | 67.15 | #454 |
| 4 | Llama 3.1 70B Instruct | 66.51 | #561 |
| 5 | GPT-4o Mini | 59.95 | #599 |
| 6 | Llama 3 70B Instruct | 53.77 | #634 |
| 7 | Qwen 2.5 7B Instruct | 45.32 | #862 |
| 8 | aya-expanse-32B | 43.9 | #692 |
| 9 | Qwen 2 7B Instruct | 43.36 | #991 |
| 10 | Llama 3.1 8B Instruct | 38.8 | #1035 |
| 11 | Llama 3 8B Instruct | 36.43 | #1129 |
| 12 | aya-expanse-8B | 33.36 | #931 |
| 13 | aya-23-35B | 32.94 | #883 |
| 14 | aya-23-8B | 27.43 | #1299 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=multilingual-mmlu-bengali · How It Works · Data refreshed daily, snapshot 2026-10-20.